Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

SENTIMENT ANALYSIS

No description
by

manasa prakash

on 13 January 2016

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of SENTIMENT ANALYSIS

SENTIMENT ANALYSIS
Manasa prakash
CONTENT
The web today and the importance of analysis
APPROACH
BRAINSTORM
ELEMENTS
SA & OP !!!
Using NLP, statistics, or machine learning methods to extract, identify, or otherwise characterize the sentiment content of a collection of text.
THE WEB TODAY
THE NEED FOR ANALYSIS
STATE OF ART
LANGUAGE ISSUES
1. NEGATION- DOUBLE & BLIND
2. INTENSIFICATION
3. ACRONYMS , JARGONS AND EMOTICONS

State of Art
Approach
Data Collection
Experiments
Results
Conclusion
Marketing tool: extracted sentiment may be used to predict future trends and behavior.
Micropublic reporting.
Longitudinal Analysis: Where are the cycles and patterns in sentiment?
Root cause analysis.
The ability to interpret human emotions.
Enables advances in NLP
Improvements in machine learning accuracy.
FUTURE: Predictive analysis & Analysis of non-textual input
METHODS
ISSUES
ACCURACY
PERFORMANCE
TIME
ACCURACY
PERFORMANCE
TIME
1. Language Issues
2. Context Issues
3. Domain Issues
4. Subjectivity Issues
1. Data Volume
2. System Nature
1. Implement
2. Test
PERFORMANCE ISSUES
1. data volume: ML METHODs NOT too SCALABLE
2. system nature: sandbox environment
TIME
1. implement: resource extraction; training time
2. TEST: REAL DATA
CONTEXT ISSUES:
1. DOUBLE MEANING
2. multiple OPINION
DOMAIN ISSUES:
POSITIVE OR NEG?
SUBJECTIVITY ISSUES
1. SARCASM
2. ASPECT IMPORTANCE
**INTERPRETATION ISSUES
1. aCTUAL RESULT COULD CONTAIN DUAL EMOtiONS.
2. every opinion is subjective and complex for even human
interpretation.
3. sometimes texts are gramatically or linguistically
incorrect .
4. stength of opinion not clearly understood.
CONCLUSION

An intensity based result following a rule based system providing information on all aspects of a phrase is found to be more accurate than a general lexicon or machine learning based method.
This system giving deeper insights also took lesser time and performed better.
However, a machine learning outlook on this would be more useful for future works. For this a huge training data set that is cross-domain and culturally&linguistically varied text would be appreciated.
A more NLP approach that handles dependencies between the nature of the words may also be considered.
In conclusion, this research has high potential for further development.
RESULTS
Accuracy
AlGORITHM
DATA COLLECTION-test&RESOURCE
A combination of keyword based and NLP techniques may be combined to give better results.
Check the nature of each word and also the nature of the term before and after it.
Taking into consideration the informal nature of the text on social media:
-Emoticons
-Jargons
-Abbreviations
-Punctuation
-Unusual Casing
-Deliberate spelling errors
Computing each polarity aspect- because that's how human brains compute.
Also check the compound nature, i.e. if two or more opinions are involved.
Data is the core of any data mining problem.
The importance of meaningful data is under-stated.
It's important to use good data in 4 stages:
-Dictionary or the lexical resource used
-The data used to train on
-The data used to test on
-The data used to implement on
Different requirements for each stage.
Importance of processing and normalizing but not losing out on useful info.
Twitter: rich, open source of data.
Use it to identify emoticons and tech jargons: empirical scoring.
Adjectives:ffective Norms for English Words (ANEW- AFINN)
Tweets: Movies.
e.g. The character was written well and was beautiful.
5-7 sec less than ML
THANK YOU!!
WHY TWITTER?!
open
so much data!
expert & naive opinions
real opinion
140 char only!
formal, informal
grammar- nazis&jews!

Full transcript